Yikes.
I always raise an eyebrow when someone mentions that they "need" more than 2 vCPU's on a terminal server.
RDS and Citrix servers benefit from scale out vs scale up. More VM's with 2-4 vCPU's and 8-12Gb of memory. Sure there are outliers to that but typical office application Citrix servers hit a sweet spot in the 2 vCPU/8-12Gb range.
3/1 ratios are pretty typical.
Must love ESXtop. Check out the ESX bible here.
Interpreting esxtop Statistics
I would say between 5-10 is getting "high" for %RDY. But it really comes down to end user experience as well. Are users experiencing performance issues? Typical causes of high allocations are a lack of understanding how to properly scale a VM for the workload. Often times other problems are the root cause but Memory and vCPU are the easiest thing to simply throw at the problem. Networking and storage for example are often root causes.
Other things I would be curious about-
- Do you by chance have these in a Resource Pool?
- Are there any CPU limits set on these VM's or do you see any %MLMTD figures appearing for these VM's?
- Do you see any %CSTP figures for any of your VM's?
I just took a look at a sample of about 20 Citrix Servers on one of our hosts and we are seeing %RDY times from .0X to 1.4X values.
My suggestion would be to shutdown a few VM's, enable hot and drop a sample group down to 2vCPU's and have users test performance while you keep a watchful eye on ESXtop.
Hope this helps!